Anthropogenicfoodresourcesfosterthecoexistenceofdistinctli ... · 71 campsites and residential...
Transcript of Anthropogenicfoodresourcesfosterthecoexistenceofdistinctli ... · 71 campsites and residential...
Zurich Open Repository andArchiveUniversity of ZurichMain LibraryStrickhofstrasse 39CH-8057 Zurichwww.zora.uzh.ch
Year: 2016
Anthropogenic food resources foster the coexistence of distinct life historystrategies: year-round sedentary and migratory brown bears
Cozzi, Gabriele ; Chynoweth, Mark ; Kusak, Josip ; Coban, Emrah ; Coban, Aysegül ; Ozgul, Arpat ;Sekercioglu, Cagan H
Abstract: Plastic behavioral adaptation to human activities can result in the enhancement and estab-lishment of distinct behavioral types within a population. Such inter-individual behavioral variations,if unaccounted for, can lead to biases in our understanding of species’ feeding habits, movement pat-tern, and habitat selection. We tracked the movements of 16 adult brown bears in a small and isolatedpopulation in northeast Turkey to i) identify inter-individual behavioral variations associated with theuse of a garbage dump and ii) to examine how these variations influenced ranging patterns, movementsbehavior and habitat selection. We identified two remarkably distinct behavioral types: bears that reg-ularly visited the dump and remained sedentary year-round, and bears that never visited the dump andmigrated 165.7 ± 20.1 km (round-trip mean cumulative distance ± SE) prior to hibernation to search forfood. We demonstrated that during migratory trips, bears moved more rapidly and were less selectivein habitat choice than during the sedentary phase; during the migration phase forest cover was the onlyimportant environmental characteristic. Our results thus reinforce the growing evidence that animals’use of the landscape largely changes according to movement phase. Our study shows that anthropogenicfood resources can influence food habits, which can have cascading effects on movement patterns andhence habitat selection, ultimately resulting in the establishment of distinct behavioral types within apopulation. Identification and consideration of these behavioral types is thus fundamental for the correctimplementation of evidence-based conservation strategies at the population level.
DOI: https://doi.org/10.1111/jzo.12365
Other titles: Bear migration and movements in a human-dominated landscape
Posted at the Zurich Open Repository and Archive, University of ZurichZORA URL: https://doi.org/10.5167/uzh-124337Journal ArticleAccepted Version
Originally published at:Cozzi, Gabriele; Chynoweth, Mark; Kusak, Josip; Coban, Emrah; Coban, Aysegül; Ozgul, Arpat; Sek-ercioglu, Cagan H (2016). Anthropogenic food resources foster the coexistence of distinct life historystrategies: year-round sedentary and migratory brown bears. Journal of Zoology, 300(2):142-150.DOI: https://doi.org/10.1111/jzo.12365
1
Anthropogenic food resources foster the coexistence of distinct life history strategies: 1
year-round sedentary and migratory brown bears 2
3
Gabriele Cozzi1,Ψ, Mark Chynoweth2, Josip Kusak3, Emrah Çoban4, Ayşegül Çoban4,5, *Arpat 4
Ozgul1, *Çağan H. Şekercioğlu2,4,6 5
6
1 Department of Evolutionary Biology and Environmental Studies, Population Ecology
Research Group, Zurich University, Winterthurerstrasse 190, CH-8057 Zürich, Switzerland.
2 Department of Biology, University of Utah, 257 South 1400 East, Salt Lake City, Utah
84112, United States.
3 Biology Department, Veterinary Faculty, University of Zagreb, Heinzelova 55, 10000
Zagreb, Croatia.
4 KuzeyDoğa Society, Ortakapi Mah. Sehit Yusuf Cad. 39, 36100 Kars, Turkey.
5 Department of Parasitology, Kafkas University Institute of Health Sciences 7
Paşaçayırı, 36100 Kars, Turkey 8
6 College of Sciences, Koç University, Rumelifeneri, Sariyer 34450 Istanbul, Turkey 9
* Co-senior author
Ψ Corresponding author: Cozzi Gabriele Email: [email protected] 10
Running title: Bear migration and movements in a human-dominated landscape 11
12
13
14
2
ABSTRACT: 15
Plastic behavioral adaptation to human activities can result in the enhancement and 16
establishment of distinct behavioral types within a population. Such inter-individual 17
behavioral variations, if unaccounted for, can lead to biases in our understanding of species’ 18
feeding habits, movement pattern, and habitat selection. We tracked the movements of 16 19
adult brown bears in a small and isolated population in northeast Turkey to i) identify inter-20
individual behavioral variations associated with the use of a garbage dump and ii) to examine 21
how these variations influenced ranging patterns, movements behavior and habitat selection. 22
We identified two remarkably distinct behavioral types: bears that regularly visited the dump 23
and remained sedentary year-round, and bears that never visited the dump and migrated 165.7 24
± 20.1 km (round-trip mean cumulative distance ± SE) prior to hibernation to search for food. 25
We demonstrated that during migratory trips, bears moved more rapidly and were less 26
selective in habitat choice than during the sedentary phase; during the migration phase forest 27
cover was the only important environmental characteristic. Our results thus reinforce the 28
growing evidence that animals’ use of the landscape largely changes according to movement 29
phase. Our study shows that anthropogenic food resources can influence food habits, which 30
can have cascading effects on movement patterns and hence habitat selection, ultimately 31
resulting in the establishment of distinct behavioral types within a population. Identification 32
and consideration of these behavioral types is thus fundamental for the correct 33
implementation of evidence-based conservation strategies at the population level. 34
35
KEY WORDS: anthropogenic food resource; behavioral type; habitat selection; migration; 36
movement modes; Ursus arctos. 37
38
39
3
INTRODUCTION 40
As a result of increasing human pressure, many wildlife species live in modified and 41
fragmented landscapes (Hanski, 1999; Goudie, 2013). To cope with novel and constantly 42
changing environments, cognitively complex species may develop plastic strategies (Valeix et 43
al., 2012; Sol, Lapiedra & González-Lagos, 2013, Flack et al., 2016), which can result in the 44
establishment of distinct alternative behaviors (hereafter behavioral types) within a population 45
(Gill, Norris & Sutherland, 2001; Elfström et al., 2014). Such inter-individual variation in 46
behavioral types, if unaccounted for, can lead to biases in our understanding of species’ life 47
history traits, movement pattern and habitat selection (Elliot et al., 2014; Weimerskirch et al., 48
2015). Therefore, careful identification and consideration of observed variation in behavioral 49
types is fundamental for the correct implementation of evidence-based conservation strategies 50
at the population level. 51
Animal behavior, life history, movement patterns and habitat selection can be 52
influenced by environmental variations (Nelson, 1998; Stien et al., 2010), by changes during 53
different stages of the life cycle, such as the transition from a sedentary to a dispersing 54
movement mode (Elliot et al., 2014), and by anthropogenic activities (Ordiz et al., 2013, 55
Flack et al., 2016). For example, the access to additional food sources resulted in a 56
subpopulation of otherwise migrant white storks (Ciconia ciconia) to remain resident year-57
round (Massemin-Challet et al., 2006). Similarly, spatiotemporal variation in anthropogenic 58
food resources influenced black-tailed gull (Larus crassirostris) foraging trips and selection 59
of feeding grounds during the incubation and hatching period (Yoda et al., 2012). Changes in 60
feeding habits, movement patterns and habitat selection thus provide us with a dynamic 61
insight in the animal’s sensitivity and adaptation to anthropogenic activities and alteration of 62
the landscape. Information on movement patterns and habitat selection during long-distance 63
movements can help us further understand a species’ requirements during different stages and 64
4
under changing environmental conditions. This knowledge is necessary to model movement 65
of individuals among habitat fragments, implement evidence-based plans to create wildlife 66
corridors, and promote connectivity among populations (Palmer, Coulon & Travis, 2014; 67
Runge et al., 2014). 68
A species that shows remarkable adaptation to human-altered landscapes is the brown 69
bear (Ursus arctos). Bears are well known to complement their diet at garbage dumps, 70
campsites and residential areas. The frequent use of these human-related food resources often 71
leads to individual bears becoming ‘problem’ animals, which are frequently relocated or 72
killed by management agencies (Peirce & Van Daele 2006). The access to artificial food 73
resources has been reported to reduce bears home range size (Blanchard & Knight, 1991), 74
despite home range in wild bears is typically not directly influenced by food availability 75
(Dahle & Swenson, 2003). Human activity and disturbance further influence the 76
spatiotemporal use of resources and movement patterns (Martin et al., 2010; Ordiz et al., 77
2013). The bears’ behavioral plasticity and individual opportunistic behavior may thus result 78
in the establishment of alternative life history traits, such as alternative feeding strategies, 79
movement patterns and habitat selection among individuals with access to artificial food 80
resources. 81
The aim of this study was to investigate the effects of an anthropogenic food resource, 82
a city garbage dump, on feeding and ranging patterns of a small and isolated subpopulation of 83
bears in northeastern Turkey. In particular, we examined whether all bears used the dump to 84
the same extent or whether they exhibited distinct feeding strategies. We expected that, if 85
distinct feeding strategies were established within the population, they should be reflected in 86
distinct spatial and movement patterns. We therefore tested for differences in movement 87
patterns and movement parameters, and investigated habitat selection between quantitatively 88
distinct sections of the entire path (i.e., the chronological collection of all its GPS locations) 89
5
of each individual. The obtained information was crucial for the implementation of local 90
management interventions, as there were governmental plans for closing the dump, with 91
predicted imminent changes in the bears’ foraging strategies. Our results on habitat selection 92
have also imminent conservation implications, as they will be used to optimize the design of 93
the first wildlife corridor in Turkey, whose globally important biodiversity and wildlife 94
populations are experiencing a major conservation crisis (Şekercioğlu et al., 2011). 95
96
METHODS 97
Study Area 98
The core study area (~550 km2) was located in northeast Turkey and included the Sarıkamış 99
Forest Allahuekber Mountains National Park (hereafter SAMNP) and the surrounding 100
landscape (Fig. 1). The climate is continental, with temperate summer months during June–101
September (average monthly: 13–18°C), and cold winter months with snowfalls during 102
November–March (average monthly: -10–0°C). 103
SAMNP covers an area of 225.2 km2, but only 49.69 km2 is forested (Capitani et al., 104
2015). The remaining 278.7 km2 of forest is not protected, for a total forest cover of 328.4 105
km2 (hereafter Sarıkamış forest). Sarıkamış forest is almost exclusively composed of Scots 106
pine (Pinus sylvestris). Open pastures and arable land surround patches of forest (Fig. 1). 107
Sarıkamış forest is fragmented, and is heavily used for logging, grazing, harvesting and 108
recreation. The understory vegetation is over-exploited, with consequent food scarcity for 109
grazers (Capitani et al., 2015) and frugivorous species. Wild ungulate prey species are very 110
rare (Capitani et al., 2015). Wolves (Canis lupus) and Caucasian lynx (Lynx lynx dinniki) also 111
inhabit the study area (Capitani et al., 2015; Chynoweth, Coban & Şekercioğlu, 2015). 112
Although a viable bear population is known to occur ca. 100 km away in the Black Sea 113
6
forests (Can & Togan, 2004), no information was available on the bear population in the 114
SAMNP region prior to this study. 115
Additional fragmented patches of forest are scattered throughout the landscape 116
considerably far (> 12 km) from Sarıkamış forest (Fig. 1). Together with their surrounding 117
landscape, these forest remnants formed the extended study area of approximately 5000 km2. 118
This extended area enclosed all locations visited by the bears during long distance movements 119
outside the core study area (see below). 120
In the middle of the core study area is the city of Sarıkamış (E 42.595°, N 40.332°) 121
with a population of 18,000 inhabitants (Fig. 1). An unfenced garbage dump lies about 3 km 122
west of the outskirts of Sarıkamış and represents a year-round additional source of food (Fig. 123
1). Bears are known to visit the dump at night and feed on food scraps (pers. obs.). The 124
proportion of the bear population visiting the dump and its effects on foraging behavior, 125
movements, and demographic traits were not previously investigated. 126
127
Fieldwork and collection of GPS movement data 128
We captured and collared ten adult males and six adult females from a small and isolated 129
population in northeastern Turkey between September 2012 and June 2014. Immobilized 130
bears were fitted with GPS/GSM or GPS/Iridium radio-collars (GPS Plus, Vectronic 131
Aerospace GmbH, Germany) programmed to record one GPS location every hour. Bears were 132
monitored for a mean duration of 296 days (range: 125 –590 days). GPS acquisition rate was 133
> 90% for 15 out of 16 individuals; one collar consistently performed poorly (acquisition rate 134
≈ 50%) (Online Supplementary Material, Appendix 1). To avoid including inaccurate GPS 135
locations in the dataset, we removed all locations with a position dilution of precision (PDOP) 136
> 10 (Elliot et al., 2014). During the winter, when bears hibernate in caves or holes 137
(interquartile range: from November 23rd – December 3rd to March 6th – April 1st), the GPS 138
7
typically failed to acquire satellites; therefore, in the analyses, we only used each individual’s 139
location data collected pre and post individual hibernation date. 140
141
Identification of inter-individual variation in the use of the garbage dump 142
For each bear and for each year, we summed the number of GPS locations at the garbage 143
dump each month. We used a generalized additive mixed model (GAMM) framework to 144
investigate the relationship between month and the number of locations at the dump, while 145
allowing for potential nonlinear relationships (Wood, 2006). We entered individual gender 146
(Appendix 1) as categorical covariate, whereas we treated individual identity as random 147
intercept. This approach allowed us to identify two distinct categories. In subsequent 148
analyses, we therefore investigated and compared movement modes, movement pattern, and 149
habitats selection between these two categories. 150
151
Investigation of movement modes and sub-division in discrete movement phases 152
To investigate whether the two different categories exhibited different movement modes, we 153
fitted four competing a-priori-defined functions representing alternative movement modes to 154
the entire path of each collared bear. The four movement modes were as follow: (i) year-155
round residency, (ii) dispersal, (iii) migration, and (iv) nomadism (sensu Börger & Fryxell, 156
2012) (Fig. 2). This analytical method relies on the net squared displacement statistics 157
combined with a non-linear hierarchical modeling framework (Börger & Fryxell, 2012). 158
Appendix 2 provides a detailed mathematical and visual representation. We developed an 159
additional metric to ensure that the NSD did not assign long-distance movement modes such 160
as migration to small-scale movement patterns occurring at the local scale. For each 161
individual, we calculated the ratio between the maximum and the median of the observed net 162
displacement (ρ). In this metric the maximum net displacement for migrating individuals 163
8
should increase faster than the median, thus increasing the value of ρ. Empirical evidence 164
suggested that ρ > 5 corresponded to actual migration events; while ρ between 1.5 and 2.5 165
were typical of individuals moving at the local scale (Fig. 2, Appendix 1). 166
In a second step, we visually sub-divided the movement mode of each individual in 167
discrete movement phases: (i) sedentary, (ii) roaming, and (iii) stopover (Fig. 2). For 168
example, an individual characterized by year-round residency was assigned a sedentary phase 169
for its entire path (Fig. 2). On the other hand, the entire path of a migrating individual was 170
chronologically divided into sedentary, roaming, stopover (the final site of the migratory trip), 171
roaming, and sedentary phases (Fig. 2). Here, migration refers to a particular movement mode 172
and hence to an entire movement trajectory, and not to the actual displacement phase between 173
two distinct geographic areas, which we define as the roaming phase. We then investigated 174
differences in movement parameters and habitat selection between the three different 175
movement phases between and within the two distinct bear categories (see Calculation of 176
movement parameters and Step selection function). 177
178
Calculation of movement parameters 179
We first investigated differences in movement parameters (i.e., step length and turning 180
angles) between day and night, as bears in European human-dominated landscapes are known 181
to be predominantly nocturnal (Kaczensky et al., 2006). Only consecutive locations 1 hour 182
apart were considered. Due to the considerable differences detected between the dial periods, 183
we recalculated movement parameters for the sedentary, stopover, and roaming phases using 184
night-only data. 185
In a subsequent step, we investigated differences in step length between the two bear 186
categories and across the three movement phases using a mixed-effects model. We included 187
sex and season as additional categorical covariates, and individual as random intercept. The 188
9
inclusion of season as covariate to control for seasonal effects was due to the fact that the 189
roaming and stopover phases were highly seasonal, and thus differences in step lengths 190
between these two phases and the sedentary phase could have arisen through seasonal 191
differences instead of through genuine differences among movement phases. 192
193
Step selection function 194
We used a step selection function (SSF) framework (Fortin et al., 2005) to infer the effects of 195
landscape structures on bear movements during the sedentary, stopover and roaming phases. 196
For each phase, we pooled the data irrespective of bear category. Step selection functions 197
typically assume an exponential function of the form: 198
𝑤 𝑿 = exp(𝛽!𝑥! + 𝛽!𝑥! +⋯+ 𝛽!𝑥!)
where βi are the coefficients estimated by conditional logistic regression associated with 199
landscape variables xi. Steps with higher SSF scores w(X) are more likely to be chosen by the 200
animals (Fortin et al., 2005), and β = 0 indicates absence of selection (Forester et al., 2009). 201
For each observed step, we created a set of 10 alternative steps; the end of these steps 202
represented alternative locations that the animal could have chosen. A step is here defined as 203
the vector between two consecutive locations. Step length refers to the Euclidean distance 204
between consecutive locations. Following Fortin et al. (2005), these alternative steps were 205
created by drawing step length and turning angles from movement-phase-specific empirical 206
distributions built with data collected from the other monitored individuals (see Appendix 3 207
for more details). 208
Landscape characteristics at the observed locations were regressed against those at the 209
alternative locations. Landscape characteristics included distance to the nearest village, 210
distance to the nearest paved road, altitude, slope, aspect, and forest cover (Appendix 3). 211
Because selection partially depends on the scale at which a resource is distributed in the 212
10
landscape, a linear variable “distance to the previous location” was included in the model to 213
increase the robustness of our analysis (Forester, Im & Rathouz, 2009). We implemented a 214
two-stage approach using the TwoStepClogit package (Craiu et al., 2011) in R (The R 215
Foundation for Statistical Computing; version 3.0.3) to allow for differential habitat selection 216
responses among individuals (Fieberg et al., 2010). We removed GPS locations at the garbage 217
dump from the analysis of the effect of landscape structure on the bears’ habitat selection. 218
This was because the dump is not a feature characteristic of the entire landscape, and 219
including these locations would have resulted in an over-representation (i.e., inflated 220
selection) of the environmental variables (such as forest cover) at the dump. We followed the 221
10-fold cross-validation procedure suggested by Boyce et al. (2002) to examine model 222
performance (see Appendix 3 for more details). 223
224
RESULTS 225
Inter-individual variation in the use of the garbage dump 226
We observed two categories of individuals: bears that visited the dump (hereafter dump bears) 227
and bears that never did (hereafter wild bears). Dump bears included three females and seven 228
males; wild bears included three females and three males. Visits at the dump significantly 229
varied across months (Fedf=5.1 = 8.93, p < 0.001) but not between gender (t = 1.5, p = 0.13). 230
Visits increased towards the second half of the year (> 40% increase between March and 231
September) and peaked in late August (Fig. 3). Dump bears hibernated on average three days 232
after wild bears (November 25th and November 22nd respectively), suggesting that this life 233
history trait is not influenced by the use of the dump. We captured three dump bears in the 234
forest 5.7, 7.2 and 10.1 km from the dump, and we observed three wild bears in the vicinity 235
of, but never at, the dump (closest recorded location 0.5, 1.3, 2.0 km). We therefore 236
11
concluded that capture site locations did not explain the existence of the two observed 237
categories. 238
239
Movement modes 240
All wild bears migrated outside Sarikamiş forest. Five individuals made long-distance 241
migratory trips characterized by a maximum linear displacement from the site of capture of 36 242
– 108 km, and lasted between 23 and 72 days. One male made a shorter migratory trip with a 243
maximum linear displacement of 17 km, which lasted only 7 days (Appendix 5). Overall, the 244
mean cumulative migratory round-trip distance was 165.7 ± 20.1 km. Migratory trips 245
occurred closely prior to hibernation between the mean dates September 18th (range: August 246
29th – September 30th) and November 1st (range: October 10th – December 11th) (Fig. 3). The 247
only exception was the male that did a shorter trip of 7 days in June. His collar stopped 248
recording GPS locations on October 9th, we cannot know whether this bear may have also 249
migrated after that date. We conservatively classified one wild male as nomadic (Appendix 250
4). His collar stopped working on November 11th and we therefore don’t know whether or not 251
he had returned to Sarikamiş forest before hibernation. 252
Dump bears never migrated, with the exception of an old female that made a shorter 253
migratory trip of 27 km that lasted 13 days (Appendix 5). Given the short duration of this trip, 254
we cannot exclude that this was a prospecting trip rather than real migration. The same 255
applies to the wild male that made a short trip of seven days. 256
257
Movement parameters 258
The distribution of step lengths and turning angles varied considerably between daytime and 259
nighttime (Fig. 4a,b). In particular, during the day, bears were characterized by turning angles 260
close to 180° and short steps (mean ± SE = 263 ± 5 m), which are typical of stationary (i.e., 261
12
resting) or small-scale searching (e.g. feeding) behavior. To the contrary, at night, their 262
movement pattern was more directional with turning angles close to 0° and displacements 263
occurred at a quicker pace (mean ± SE = 535 ± 5 m). At night, the distribution of step lengths 264
and turning angles showed more consistent patterns across the three movement phases (Fig. 265
4c,d). Nevertheless, steps during the roaming phase appeared longer and more directional. 266
We did not detect differences in overall nighttime step lengths between wild and dump 267
bears (F1,12 = 0.4, p = 0.52) nor between gender (F1,12 = 1.48, p = 0.25). Step length differed 268
significantly among movement phases (F2,34275 = 511.8, p < 0.001). Irrespective of behavior 269
and sex, steps during the roaming phase were twice as long (predicted mean step length 940 270
m) than steps during the sedentary (439 m) and stopover (420 m) phase (Fig. 4e). We detected 271
a significant seasonal effect (F2,34275 = 62.7, p < 0.001) with a tendency towards shorter steps 272
late in the season, suggesting that the difference between the sedentary and roaming phases 273
did not depend on seasonal factors, but rather on genuinely different patterns during the 274
movement phases. If this was not the case, a reduction, rather than an increase, in step length 275
during the roaming phase should have been observed. 276
277
Habitat selection 278
Results are based on data from wild and dump bears for the sedentary phase and on data from 279
wild bears for the roaming and stopover phase. At the population level, bears appeared to be 280
less ‘selective’ in their habitat choice during the roaming phase than they were during the 281
sedentary and stopover phase. During roaming, out of the six landscape variables, only the β 282
coefficient for forest had a value greater than 2 SE from 0 (Tab. 1). This indicates a 283
significant association between forest and the bears’ chosen paths. Nevertheless, we observed 284
high inter-individual variation for forest selection (Table 1). 285
13
During the sedentary phase, forest, slope and distance to roads significantly influenced 286
the animals’ step selection (Tab. 1). The positive effect of slope and forest suggests that, at 287
the population level, animals sought forested locations and steeper slopes. The negative 288
relationship with distance to roads indicates that locations far away from roads were less 289
likely to be chosen. During the stopover phase bears preferred forest and locations far from 290
villages (Tab. 1). Based on 10-fold cross-validation procedure, our models provided excellent 291
fit for the sedentary phase (rs = 0.95) and only moderate for the stopover (rs = 0.23) and 292
roaming (rs = 0.12) phase. 293
294
DISCUSSION 295
We defined two categories of bears based on high-resolution GPS data from 16 adult 296
individuals: dump bears (i.e. bears that regularly visited a garbage dump) and wild bears (i.e. 297
bears that never did). Substantial differences in movement patterns between dump and wild 298
bears allowed us to identify two distinct behavioral types. While dump bears were 299
characterized by year-round residency, wild bears undertook migratory round-trips > 100 km. 300
Our results thus showed that differences in life history traits within the study population were 301
associated with the exploitation of a human-related food source. To the best of our 302
knowledge, such behavioral dichotomy within a population of brown bears has never been 303
reported; and only a few cases are known for black bears (Ursus americanus) (Noyce & 304
Garshelis, 2011; Liley & Walker, 2015). Extreme variation in migratory behavior have been 305
shown to have direct energetic and fitness consequences (Weimerskirch et al., 2015, Flack et 306
al., 2016). Investigation of differences in key demographic parameters such as survival and 307
reproductive rate between the two behavioral types is therefore required to better understand 308
the population dynamics of the study system. 309
14
Migration is conceivably linked to a seasonal availability of resources, such as food 310
and mates (Dingle, 2014). Because migratory trips occurred right before hibernation and 311
because direct field investigation of the vegetation at migration stopover sites revealed a high 312
prevalence of oak forest (Quercus spp.) (as opposed to Sarikamiş forest which was entirely 313
composed of Scots pines; cfr. Appendix 5), we deduced that hyperphagia before the winter 314
drove the observed patterns (Noyce & Garshelis, 2011, Seger et al., 2013). This hypothesis 315
was further corroborated by the fact that only wild bears (i.e. those bears that did not use the 316
additional food resources provided by the city garbage dump) migrated. Since migratory trips 317
occurred between September and November, we excluded mating activities (May – July) as 318
an alternative driver for the observed movement patterns. We found no comparable study 319
describing similar food-related migratory movements immediately before hibernation in 320
brown bears. Additionally, while long-distance movements of bears are typically associated 321
with dispersal or translocation events (Liley & Walker, 2015) the observed distances covered 322
by migrating wild bears were remarkable. Our findings thus add valuable information to the 323
life history of the species and a new spatiotemporal dimension to its management and to 324
conservation efforts. 325
The identification of two behavioral types and information on ranging patterns have 326
far-reaching implications for the regional management and conservation of the species. First, 327
the observed long-distance movements showed that bears living in the SAMNP are 328
potentially connected with the larger bear population of the Black Sea mountains and Georgia 329
(Can & Togan, 2004). Our data also provided further support for the ongoing efforts to create 330
Turkey’s first wildlife forested corridor, with the goal of enhancing connectivity between the 331
SAMNP and wildlife populations in the Black Sea and Georgian forests. Second, the natural 332
resources of Sarıkamış forest may not be sufficient to sustain the local bear population 333
throughout the entire year. Bears had to migrate to find food outside the core study area or to 334
15
supplement their diet with anthropogenic food resources. Any intervention that would limit 335
either option could have severe consequences at the subpopulation level. Third, following a 336
governmental plan, the city garbage dump will be closed in the near future. We hypothesized 337
three scenarios: 1) dump bears die following malnutrition before hibernation, 2) they resume 338
the migratory behavior observed in forest bears, or 3) they seek food in the Sarıkamış city and 339
nearby villages. Given the bears’ ability to exploit anthropogenic food resources (Elfström et 340
al., 2012), we anticipate the third scenario, at least in the short term, which is likely to 341
increase the interactions and existing conflicts with people (Chynoweth et al., 2016). To limit 342
interactions and avoid fatalities, the closure of the dump should therefore be coupled with the 343
measures such as the use of bear-proof bins and daily removal of household leftovers 344
(Robbins, Schwartz & Felicetti, 2004). In the long term, after the dump closure, the 345
persistence of the bear population of the Sarıkamış forest will depend on the bears’ migratory 346
possibility. Conservation efforts should therefore aim to secure and facilitate their migratory 347
movements to the foraging grounds prior to hibernation. Given the population-level selection 348
for forested habitat, this can be achieved through the reforestation of the proposed wildlife 349
corridor and should be accompanied by education efforts to enhance bear acceptance by the 350
local population along the observed migratory route. 351
We also demonstrated that animals’ movements and use of the surrounding landscape 352
largely depend on their movement phase. Our study thus provides further evidence that the 353
source of the data used to model animals’ habitat selection is as important as the type of 354
predictor environmental variables considered (Zeller, McGarigal & Whiteley, 2012; Elliot et 355
al., 2014). We showed that during the roaming phase bears were less selective in their habitat 356
choice compared to the sedentary phase. Differences in habitat selection between resident and 357
roaming individuals (in the specific case of dispersers) have been reported for other species 358
(Elliot et al., 2014; Killeen et al., 2014). While during the sedentary phase individuals may 359
16
select habitats based on the ‘known’ distribution of food, shelter, and mates, during the 360
roaming phase they are more naïve to the landscape matrix they move through. Forest 361
appeared, however, to be equally important in each phase. Being a prominent landscape 362
feature, forest can be easily recognized in the distance during migratory trips through 363
unknown landscapes, and actively selected for. The selection of locations closer to roads 364
during the sedentary phase around Sarikamiş forest may be due to the presence of additional 365
food resources deriving from intensive picnic activities (pers. obs.), but further investigation 366
is necessary. Including non-linear responses of distance to roads, but also of distance to 367
villages and elevation, could help further understand mechanisms of habitat selection. We 368
caution for over interpretation of the results for the roaming and stopover phase due to the 369
moderate model performance (see Appendix 3 for further considerations). 370
To summarize, we showed that the availability of a human-related source of food can 371
cause a behavioral dichotomy among individuals of a confined population. This inter-372
individual variation is manifested in alternative feeding habits, movement pattern and 373
selection of different habitat types. Therefore, identification and consideration of observed 374
variation in behavioral types is fundamental for the correct implementation of evidence-based 375
conservation strategies. Failures to detect such differences could result in the erroneous 376
allocation of limited conservation resources, such as setting aside portions of land 377
characterized by landscape features that are critical to only particular behavioral types 378
(Simberloff et al., 1992; Beier & Noss, 2008). Finally, because most research on brown bears 379
has been carried out in northern Europe and North America, this work in Turkey increases our 380
understanding of the species living under considerably different environmental, ecological, 381
and social conditions. Empirical evidence from this work thus adds valuable information for 382
the implementation of management and conservation strategies of bears in Turkey but also 383
worldwide. 384
17
ACKNOWLEDGEMENTS 385
We thank the General Directorate of Nature Conservation and National Parks and Forestry 386
General Directorate of Turkey’s Ministry of Forestry and Water Affairs for permitting our 387
research. For their support, we thank the Forschungskredit der Universität Zürich, Claraz 388
Foundation, Ashoka Fellowship, Bosporus University, Christensen Fund, Fondation Segré, 389
National Geographic Society, National Science Foundation, UNDP Small Grants Program, 390
University of Utah, Whitley Fund and KuzeyDoğa and Nature Türkiye Foundation donors, 391
especially Bilge Bahar, Seha İşmen, Lin Lougheed and Batubay Özkan. We are grateful to the 392
KuzeyDoğa staff and volunteers for their efforts through the years and to the people of 393
Erzurum, Kars and Sarıkamış for their hospitality. 394
395
REFERENCES 396
Bastille Rousseau, G., Fortin, D., Dussault, C., Courtois, R. & Ouellet, J.-P. (2011). Foraging 397
strategies by omnivores: are black bears actively searching for ungulate neonates or are 398
they simply opportunistic predators? Ecography 34, 588–596. 399
Beier, P. & Noss, R.F. (2008). Do Habitat Corridors Provide Connectivity? Conservation 400
Biology 12, 1241–1252. 401
Blanchard, B.M. & Knight, R.R. (1991). Movements of yellowstone grizzly bears. Biological 402
Conservation 58, 41–67. 403
Börger, L. & Fryxell, J.M. (2012). Quantifying individual differences in dispersal using net 404
squared displacement. In Dispersal Ecology and Evolution. Clobert, J., Baguette, M., 405
Benton, T.G. & Bullock, J.M. (Eds). Oxford University Press. 406
Boyce, M.S., Vernier, P.R., Nielsen S.C. & Schmiegelow, F.K.A. (2002). Evaluating resource 407
selection functions. Ecological Modelling 157, 281-300. 408
Can, Ö. & Togan, İ. (2004). Status and management of brown bears in Turkey. Ursus 15, 48–409
53. 410
Capitani, C., Chynoweth, M., Kusak, J. & Coban, E. & Şekercioğlu, Ç. H. (2015). Wolf diet 411
in an agricultural landscape of north-eastern Turkey. Mammalia, in press. Availabe online 412
Chynoweth, M., Coban, E. & Şekercioğlu, C.H. (2015). Conservation of a new breeding 413
population of Caucasian lynx (Lynx lynx dinniki) in eastern Turkey 2. Turkish Journal of 414
Zoology 39, 541–543. 415
18
Chynoweth, M.W., Çoban, E., Altın, Ç., & Şekercioğlu, Ç. H. (2016). Human-wildlife 416
conflict as a barrier to large carnivore management and conservation in Turkey. Turkish 417
Journal of Zoology 40, in press. Available online 418
419
Craiu, R.V., Duchesne, T., Fortin, D. & Baillargeon, S. (2011). Conditional logistic 420
regression with longitudinal follow-up and individual-level random coefficients: A stable 421
and efficient two-step estimation method. Journal of Computational and Graphical 422
Statistics 20, 767–784. 423
Dahle, B. & Swenson, J.E. (2003). Home ranges in adult Scandinavian brown bears (Ursus 424
arctos): effect of mass, sex, reproductive category, population density and habitat type. 425
Journal of Zoology 260, 329–335. 426
Dingle, H. (2014). Migration. The Biology of Livfe on the Move. 2. edn. Oxford: Oxford 427
University Press. 428
Elfström, M., Zedrosser, A., Jerina, K., Støen, O.-G., Kindberg, J., Budic, L., Jonozovič, M. 429
& Swenson, J.E. (2014). Does despotic behavior or food search explain the occurrence of 430
problem brown bears in Europe? Journal of Wildlife Management. 78, 881–893. 431
Elfström, M., Zedrosser, A., Støen, O.-G. & Swenson, J.E. (2012). Ultimate and proximate 432
mechanisms underlying the occurrence of bears close to human settlements: review and 433
management implications. Mammal Review 44, 5–18. 434
Elliot, N.B., Cushman, S.A., Macdonald, D.W. & Loveridge, A.J. (2014). The devil is in the 435
dispersers: predictions of landscape connectivity change with demography. Journal of 436
Applied Ecology 51, 1169–1178. 437
Fieberg, J., Matthiopoulos, J., Hebblewhite, M., Boyce, M.S. & Frair, J.L. (2010). Correlation 438
and studies of habitat selection: problem, red herring or opportunity? Philosophical 439
Transaction of the Royal Society B 365, 2233–2244. 440
Flack, A., Fiedler, W., Blas, J., Pokrovsky, I., Kaatz, M., Mitropolosky, M., Aghababyan, K., 441
Fakriadis, I., Makrigianni, E., Jerzak, L., Azafzaf, H., Feltrup-Azafzaf, C., Rotics, S., 442
Mokotjomela, T., NAthan, R., Wikelski, M. (2016). Costs of migratory decisions: A 443
comparison across eight white stork populations. Science Advances 2, 1–7. 444
Forester, J.D., Im, H.K. & Rathouz, P.J. (2009). Accounting for animal movement in 445
estimation of resource selection functions: sampling and data analysis. Ecology 90, 3554–446
3565. 447
Fortin, D., Beyer, H.L., Boyce, M.S., Smith, D.W., Duchesne, T. & Mao, J.S. (2005). Wolves 448
influence elk movements: Behavior shapes a trophic cascade in Yellowstone National 449
Park. Ecology 86, 1320–1330. 450
Gill, J.A., Norris, K. & Sutherland, W.J. (2001). Why behavioural responses may not reflect 451
the population consequences of human disturbance. Biological Conservation 97, 265–452
268. 453
Goudie, A.S. (2013). The Human Impact on the Natural Environment: Past, Present, and 454
Future. Seventh edition. Oxford: John Wiley & Sons. 455
19
Hanski, I.A. (1999). Metapopulation ecology. Oxford: Oxford University Press. 456
Kaczensky, P., Huber, D., Knauer, F., Roth, H., Wagner, A. & Kusak, J. (2006). Activity 457
patterns of brown bears (Ursus arctos) in Slovenia and Croatia. Journal of Zoology 269, 458
474–485. 459
Killeen, J., Thurfjell, H., Ciuti, S., Paton, D., Musiani, M. & Boyce, M.S. (2014). Habitat 460
selection during ungulate dispersal and exploratory movement at broad and fine scale 461
with implications for conservation management. Movement Ecology 2, 15–28. 462
Liley, S.G. & Walker, R.N. (2015). Extreme movement by an American black bear in New 463
Mexico and Colorado. Ursus 26, 1–6. 464
Martin, J., Basille, M., Van Moorter, B., Kindberg, J., Allaine, D. & Swenson, J.E. (2010). 465
Coping with human disturbance: spatial and temporal tactics of the brown bear (Ursus 466
arctos). Canadian Journal of Zoology 88, 875–883. 467
Massemin-Challet, S., Gendner, J.-P., Samtmann, S., Pichegru, L., Wulgue, A. & Le Maho, 468
Y. (2006). The effect of migration strategy and food availability on White Stork Ciconia 469
ciconia breeding success. Ibis 148, 503–508. 470
Nelson, M.E. 1998. Development of migratory behavior in northern white-tailed deer. 471
Canadian Journal of Zoology 76, 426–432. 472
Noyce, K.V. & Garshelis, D.L. (2011). Seasonal migrations of black bears (Ursus 473
americanus): causes and consequences. Behavioral Ecology and Sociobiology 65, 823–474
835. 475
Ordiz, A., Støen, O.-G., Sæbø, S., Sahlén, V., Pedersen, B.E., Kindberg, J. & Swenson, J.E. 476
(2013). Lasting behavioural responses of brown bears to experimental encounters with 477
humans. Journal of Applied Ecology 50, 306–314. 478
Palmer, S.C.F., Coulon, A. & Travis, J.M.J. (2014). Inter-individual variability in dispersal 479
behaviours impacts connectivity estimates. Oikos 123, 923–932. 480
Peirce, K. N. & Van Daele, L. J. (2006). Use of a garbage dump by brown bears in 481
Dillingham, Alaska. Ursus 17, 165–177. 482
Robbins, C.T., Schwartz, C.C. & Felicetti, L.A. (2004). Nutritional ecology of ursids: a 483
review of newer methods and management implications. Ursus 15, 161–171. 484
Runge, C.A., Martin, T.G., Possingham, H.P., Willis, S.G. & Fuller, R.A. (2014). Conserving 485
mobile species. Frontiers in Ecology and the Environment 12, 395–402. 486
Seger, R.L., Servello, F.A., Cross, R.A. & Keisler, D.H. (2013). Body mass and mast 487
abundance influence foraging ecology of the American black bear (Ursus americanus) in 488
Maine. Canadian Journal of Zoology 91, 512–522. 489
Simberloff, D., Farr, J.A., Cox, J. & Mehlman, D.W. (1992). Movement corridors: 490
conservation bargains or poor investments? Conservation Biology 6, 493–504. 491
Sol, D., Lapiedra, O. & González-Lagos, C. (2013). Behavioural adjustments for a life in the 492
20
city. Animal Behavior 85, 1101–1112. 493
Şekercioğlu, Ç.H., Anderson, S., Akçay, E., Bilgin, R., Can, Ö., Semiz, G., Tavşanoğlu, C., 494
Yokeş, M., Soyumert, A., İpekdal, K., Sağlam, İ.K., Yücel, M. & Dalfes, H.N. (2011). 495
Turkey's globally important biodiversity in crisis. Biological Conservation 144, 18–18. 496
Stien, A.,Loe, L.E., Mysterud, A., Severinsen, T., Kohler, J., & Langvatn, R. (2010). Icing 497
events trigger range displacement in a high-arctic ungulate. Ecology 91, 915-920. 498
Valeix, M., Hemson, G., Loveridge, A.J., Mills, G. & Macdonald, D.W. (2012). Behavioural 499
adjustments of a large carnivore to access secondary prey in a human-dominated 500
landscape. Journal of Applied Ecology 49, 73–81. 501
Weimerskirch, H., Delord, K., Guitteaud, A., Phillips, R.A. & Pinet, P. (2015). Extreme 502
variation in migration strategies between and within wandering albatross populations 503
during their sabbatical year, and their fitness consequences. Sci Rep 5, 8853–8860. 504
Wood, S. (2006). Generalized Additive Models. Boca Raton: CRC Press. 505
Yoda, K., Tomita, N., Mizutani, Y., Narita, A. & Niizuma, Y. (2012). Spatio-temporal 506
responses of black-tailed gulls to natural and anthropogenic food resources. Marine 507
Ecology Progress Series 466, 249–259. 508
Zeller, K.A., McGarigal, K. & Whiteley, A.R. (2012). Estimating landscape resistance to 509
movement: a review. Landscape Ecology 27, 777–797. 510
511
512
513
514
515
516
517
518
519
520
521
522
523
21
FIGURES AND TABLES 524
Figure 1: The study area in northeastern Turkey. The dotted ellipse represents the core study 525
area including Sarikamiş forest and surrounding pastures. The extended study area enclosed 526
all locations visited by the bears during migratory trips outside the core study area. GPS 527
relocations of one migrating bear (wild bear) and one bear resident all-year-round (dump 528
bear) are shown as an example. 529
Black&sea&
Turkey&
Mediterranean&sea&
Georgia&
Study&area&
22
Figure 2: Characterization of movement trajectories by means of the net squared displacement 530
approach. A) Graphical representation of four alternative movement modes (underlined) and 531
sub-division in three discrete movement phases 532
(italic) (modified from Bunnefeld et a. 2011). 533
For instance, a migratory movement mode is 534
characterized by two sedentary phases, two 535
roaming phases, and one (or more) stopover 536
phase. A resident mode is characterized by a 537
sedentary phase throughout the entire 538
movement path. B) Observed trajectory 539
corresponding to a migratory movement mode 540
(left panel) and its corresponding ND (right). 541
Discrete movement phases are shown. Each dot 542
in the left panel represent a GPS location 543
collected at hourly intervals; lines connect 544
consecutive locations. In the right panel, 545
horizontal net displacement sections represent 546
the hibernation period C) Observed trajectory 547
corresponding to a resident movement mode 548
(left) and its corresponding ND (right). ρ are 549
given for both movement modes: a low value 550
indicates small-scale movements (see main text 551
for further details). 552
553
554
Nomadic
Dispersing
Resident
Migrating
Sedentary: 1
Roaming: 2
Stopover: 3
1 1
11
1
1
2
22
2
2
3
12
Time (days)
NS
D (
km
2)
A)!
Longitude
Latitu
de
265000 275000 285000
4460000
4470000
4480000
4490000
4500000
4510000
4520000 A1)
Stopover
Sedentary
Roaming
ρ=17.8
Ellapsed days
Dis
tance fro
m o
rigin
(m
)
0 50 100 150 200 250 300
010000
20000
30000
40000
50000
60000
A2)
Stopover
Sedentary
Roaming
La
titu
de!
ρ=17.8!
0!!!!!!!!!!!!!!10!!!!!!!!!!!!20!!!!!!!!!!!!!30!!!!!!!!!!!!40!!!!!!!!!!!!!50!!!!!!!!!!!!60!!!!!
0!!!!!!!!!!!100!!!!!!!!!200!!!!!!!!300!265000!!!!!!!!!!!!!!!!!!!285000!
4460000!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!4500000!!!!!!!!!!!!!!!!!!!!!!!!!!!
Sedentary
Sedentary
Roaming
Stopover
Roaming
Stopover
B)!
ND
(km
)
Longitude
La
titu
de
270000 290000 310000
44
60
00
04
48
00
00
45
00
00
04
52
00
00
45
40
00
04
56
00
00
B1)
Sedentary
ρ=1.7
Ellapsed days
Dis
tan
ce
fro
m o
rig
in (
m)
0 100 200 300 400 500
01
00
00
20
00
03
00
00
40
00
05
00
00
60
00
0
B2)
Sedentary
Longitude! Days!
La
titu
de!
ρ=1.7!
0!!!!!!!!!!!!!!200!!!!!!!!!!!!400!270000!!!!!!!!!!!!!!!!!300000!
0!!!!!!!!!!!!!!10!!!!!!!!!!!!20!!!!!!!!!!!!!30!!!!!!!!!!!!40!!!!!!!!!!!!!50!!!!!!!!!!!!60!!!!!
4460000!!!!!!!!!!!!!!!!!!!!!!!4500000!!!!!!!!!!!!!!!!!!!!!!!4540000!!!!!
Sedentary Sedentary
C)!
ND
(km
)
23
Figure 3: Presence likelihood at the city garbage dump across a year. Confidence intervals are 555
shown in grey. Bears that visited the dump (dump bears) remained resident year-round. On 556
the other side, bears that never visited the dump (wild bears) migrated before hibernation; the 557
hatched area shows the migration period. 558
559
560
561
562
563
564
565
566
567
568
Month of year
Pre
sence lik
elih
ood a
t dum
psite
2 4 6 8 10 12
0.0
0.2
0.4
0.6
0.8
1.0
Month of year
Pre
se
nce
lik
elih
oo
d a
t d
um
p
24
Figure 4: Changes in movement parameters (i.e. step length and turning angle) between day 569
and night (A, B), and among different movement phases at night (C-E). In panels C and D 570
sedentary phase data of dump and wild bears are pooled. Panel E shows differences between 571
dump and wild bears (the latter do not have a migratory and stopover phase). 572
573
Turning angle (degrees)
De
nsity
−180 −90 0 90 180
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40C)
Step length (m)
0 500 1000 1500 2000
0.0000
0.0005
0.0010
0.0015
0.0020
0.0025SedentaryStopoverRoaming
D)
Ste
p le
ng
th (
m)
Sedentary Stopover Roaming
200
400
600
800
1000
1200
1400
E)
Dump bears Wild bears
Turning angle (degrees)
De
nsity
−180 −90 0 90 180
0.10
0.15
0.20
0.25
0.30
A)
Step length (m)
0 500 1000 1500 2000
0.000
0.005
0.010
0.015
0.020 DayNight
B)
25
Table 1: Population level coefficients, estimated standard errors and variance of random 574
coefficients from a mixed conditional logistic regression of movement steps on six different 575
environmental variables during the sedentary, stopover, and roaming phases. Asterisks 576
indicate values significantly different from 0. For the sedentary phase data from dump and 577
wild bears were pooled. 578
Sedentaryphase Stopoverphase Roamingphase
beta SE Var beta SE Var beta SE Var
Distanceto
previous-0.000326 0.00019 5.5e-07 -0.000376 0.00038 8.5e-07 0.000047 0.00012 8.4e-08
Altitude 0.000828 0.00069 5.1e-06 -0.000176 0.00087 2.8e-06 0.001629 0.00136 1.1e-05
Slope 0.012802* 0.00273 5.1e-05 -0.002962 0.0028 2.1e-06 0.008096 0.00877 2.9e-04
Aspect -0.000156 0.00014 4.7e-08 0.000536 0.00035 1.9e-07 -0.000019 0.0004 1.5e-07
Forest 0.262084* 0.09703 1.0e-01 0.393145* 0.16658 8.7e-02 0.368015* 0.13588 1.9e-02
Distanceto
village0.000154 0.00013 2.2e-07 0.000291* 0.00012 3.3e-08 -0.000055 0.00008 1.1e-08
Distanceto
road-0.00024* 0.00009 8.8e-08 0.000001 0.00012 2.9e-08 0.000025 0.00006 1.4e-09
579
580
581
582
583